This invention generally relates to the fabrication of micro-electromechanical switches (MEMS), and more particularly, to the manufacture of MEMS which can be integrated into current state of the art semiconductor fabrication processes.
Switching operations are a fundamental part of many electrical, mechanical and electromechanical applications. MEM switches have drawn considerable interest over the last few years. Products using MEMS technology are widespread in biomedical, aerospace, and communications systems.
Conventional MEMS typically utilize cantilever switches, membrane switches, and tunable capacitor structures as described, e.g., in U.S. Pat. No. 6,160,230 to McMillan et al., U.S. Pat. No. 6,143,997 to Feng et al., U.S. Pat. No. 5,970,315 to Carley et al., and U.S. Pat. No. 5,880,921 to Tham et al. MEMS devices are manufactured using micro-electromechanical techniques and are used to control electrical, mechanical or optical signal flows. Such devices, however, present many problems because their structure and innate material properties require them to be manufactured in lines that are separate from conventional semiconductor processing. This is usually due to the different materials and processes which are not compatible and, therefore, which cannot be integrated in standard semiconductor fabrication processes.
The use of materials typically used in the manufacture of MEMS, such as gold, pose obvious integration problems for integrating devices directly to on-chip applications. Even the use of polysilicon, which is widely found in the literature, poses problems due to the temperature cycles and the usual segregation of front-end of the line (FEOL) tools where the actual semiconductor devices are fabricated and the back-end of the line (BEOL) where interconnect metals are processed. Typically, the two sets are not allowed to have process crossovers from one to the other in order to prevent metallic contamination of the active devices. It is therefore unlikely to see polysilicon deposition in the back-end of the line.
Most existing processes suffer from a serious drawback in that by using standard metalization, no encapsulation is provided to protect the metal. Moreover, more than one substrate is used, oftentimes bonded together, with corresponding inherent disadvantages.
Other existing techniques only provide switching capabilities at the top of the structure, making it unlikely that integration can be achieved at all levels, as will be described hereinafter in the present invention.
Accordingly, there is a need for a process that is capable of providing MEMS devices using established BEOL materials coupled to processing that can be fully integrated so that these devices can be manufactured either in conjunction with or as an add-on module to the conventional BEOL or interconnect levels.
In order to gain a better understanding of the present invention, a conventional MEM switch will now be described with reference to FIG. 1, which shows a cross-section view of a MEM switch having both ends of a deformable beam 1 anchored in dielectric 4. The lowest level consists of a dielectric material 5 containing conductive elements 2, 2a, and 3 which will be used subsequently to connect or form the various electrical portions of the device. The conductors referenced by numerals 2 and 2a are used to provide an operating potential that cause the beam to deform. Conductor 3, which conducts a signal, is in turn connected to the beam when it is in operation.
FIG. 2 shows a planar view of the same prior art MEM device of FIG. 1. In a typical implementation, deformable beam 1 is formed by polysilicon over dielectric 4, e.g., SiO2, and the surrounding material is etched away leaving a raised structure, i.e., the beam suspended above the conductors that were previously formed or which, themselves, are made of polysilicon. Then the device is subjected to electroless plating, usually of gold, that adheres to the polysilicon forming the conductive elements 1, 2, 2a and 3. The switch is operated by providing a potential difference between the beam and electrodes 2 and 2a. This voltage generates an electrostatic attraction which pulls beam 1 in contact with electrode 3, thus closing the switch.
One should note that these are all typically raised structures having a large topography when compared to conventional semiconductor devices. This in itself makes them virtually impossible to integrate into the semiconductor chip fabrication process. These devices are a typically made using surface micro-machining techniques which include building on photoresist or building on a substrate, such as silicon, and then removing a portion of the substrate under the device from the backside of the substrate, again precluding integration with standard semiconductor processing.
FIG. 3 illustrates a cross-section view of another version of a conventional MEM switch, wherein only one end of the beam 1 is anchored within the dielectric 4. All the other parts perform as described in FIG. 1. The same applies with regard to FIG. 4, illustrating a top-down view of the corresponding device illustrated in FIG. 3. In the latter case, the switch is operated by applying a voltage between beam 1 and control electrode 2. This causes the beam to be pulled down into contact with the signal electrode 3. When the voltage is dropped, beam 1 returns to its original position.
Typically, the gap between the beam and the control electrode substantially determines the voltage required to pull down the beam. Most literature describes devices having gaps ranging from 1 to several micrometers. These gaps are large and the voltage required is therefore higher than would be desired for most consumer applications. Reported activation voltages range from around 30 to 75 volts. This is far too high for applications like cell phones which typically operate between 3 to 5 volts. The structure of the present invention operates with gaps ranging from 200 angstroms to several thousand angstroms, producing switches having an activation voltage below 5 volts.
The aforementioned illustrative switch configurations are only some of many possible structures which are known in the art. It is worth noting that MEM switches may also be configured in an arrangement of multiple beams wired in a variety of combinations.
Stiction is of primary concern in MEMS devices. Stiction is defined as two or more surface making contact that will not release without causing some damage to the device. Impingement is a major cause of this phenomena. The present invention addresses this problem in at least one embodiment by providing an air gap 200 when the switch is closed, as will be shown in detail with reference to FIG. 19A. Surface tension is also believed to be another major cause of stiction. That explains why the present invention utilizes dry etches and processes for the release of the moving parts and subsequent processing.
Accordingly, it is an object of the invention to build MEM switches and other similar structures which are fully integrated within CMOS , bipolar or BiCMOS wafers.
It is another object to manufacture MEM switches and other similar structures with a modified damascene process.
It is a further object to build MEM switches and other similar structures utilizing copper encapsulated in a barrier material to protect the metal.
It is yet another object to ensure that the encapsulation can be integrated into BEOL copper at a temperature compatible with such a process.
These and other objects are addressed by the present invention by providing a method of fabricating MEMS switches integrated with conventional semiconductor interconnect levels, using compatible processes and materials.
The invention described herein provides a method of fabricating a capacitive switch adaptable to produce various configurations used for contact switching and/or metal-dielectric-metal switches.
In a preferred embodiment, the process starts with a copper damascene interconnect layer made of metal conductors inlaid within a dielectric. All or portions of the copper interconnects are recessed to a degree sufficient to provide a capacitive air gap when the switch is closed and, additionally, to provide a space for a protective layer of Ta/TaN, if so desired. The metal structures that are defined within the area specified by the switch act as actuator electrodes to pull down a moveable conductive beam providing one or more paths for the switching signals to traverse. The advantage of such air gap is that the dielectric (air) is not subject to charge storage or trapping that causes reliability and voltage drift problems.
The present invention can be embodied using any number of dielectric materials, such as silicon dioxide or silicon nitride, all of which may advantageously be placed between the lower electrodes and the deformable beam of the capacitive switch.
It is worth noting that instead of recessing the electrodes to provide a gap, one may just add dielectric around and/or on the electrode. A second dielectric layer is then deposited to the desired thickness of the gap formed between the lower electrodes and the moveable beam. Vias are the formed through the second dielectric layer to provide connections between the aforementioned metal interconnect layer and the next metal layer which includes the movable beam. The layer containing the vias is then patterned and etched to provide a cavity area which contains the lower activation electrodes as well as the signal paths. The cavity is then back-filled with sacrificial release material, preferably SiLK, which is a semiconductor dielectric in the form of a polymer resin consisting of gamma-butyrolactone, B-staged polymer and mesitylene. This release material is then made substantially planar, preferably, by way of chemical-mechanical polish (CMP) to align it with the top of the dielectric. It is also possible to planarize the release material by re-imaging the same mask that was used to define the cavity area using the opposite polarity photoresist and then etching the release material from the upper surface to make it substantially co-planar with the dielectric in which the cavity was formed. This provides a planar surface upon which the beam layer can be constructed.
The layer containing the movable beam is similar to other interconnect layers typically used for the fabrication of standard CMOS devices. The beam structure preferably extends, partially or completely, across the cavity area previously defined, such that when the switch is in operation, it provides the desired electrical connections. It should also be noted that holes may be incorporated into the moveable beam. These holes are initially filled with the same dielectric that forms the layer containing the beam. These areas of dielectric within the beam are then etched out to provide greater access to the sacrificial material under the beam and, thus, aiding the release process. Next, the pattern of the cavity area or some subset of the pattern is defined and the dielectric material surrounding the beam is removed, including areas through the holes in the beam, providing access to the underlying release material. At this point the released material can be removed. The switch, suspended over the cavity area is now fully functional.